Development of Machine Learning Models to Predict Tumor Endoprosthesis Survival.

Area of Science:

• Orthopedic oncology
• Biomedical engineering
• Machine learning in healthcare

Background:

• Endoprosthetic reconstruction is a key limb salvage technique for malignant bone tumors.
• Implant failure is a frequent complication, lacking reliable patient-specific survival predictions.
• Accurate prognostication is crucial for managing patient expectations and guiding treatment.

Purpose of the Study:

• To evaluate and compare machine learning (ML) models for predicting early survival of tumor endoprosthetic implants.
• To develop patient-specific survival estimations for improved clinical decision-making.

Main Methods:

• A retrospective analysis of 138 patients undergoing endoprosthetic reconstruction.
• XGBoost, random forest, decision tree, and logistic regression models were trained and tested.
• Features included age, sex, BMI, diagnosis, location, resection length, and number of surgeries; outcomes were 12, 24, and 36-month implant survival.

Main Results:

• The random forest model demonstrated superior performance across all time points (12, 24, 36 months).
• Achieved high AUC (0.96 at 12 months) and accuracy (0.92 at 12 months).
• Resection length was the most important feature at 12 months, while age was critical at 24 and 36 months.

Conclusions:

• Machine learning models offer accurate prediction of early endoprosthetic implant survival in tumor surgery.
• These are the first ML models to predict survival beyond one year and include upper extremity implants.
• The models provide enhanced patient-specific prognostication, aiding in expectation management and treatment recommendations.